TY - JOUR
T1 - A model-based objective evaluation of eye movement correction in EEG recordings
AU - Kierkels, J.J.M.
AU - Boxtel, van, G.J.M.
AU - Vogten, L.L.M.
PY - 2006
Y1 - 2006
N2 - We present a method to quantitatively and objectively compare algorithms for correction of eye movement artifacts in a simulated ongoing electroencephalographic signal (EEG). A
realistic model of the human head is used, together with eye tracker data, to generate a data set in which potentials of ocular and cerebral origin are simulated. This approach bypasses the common problem of brain-potential contaminated electro-oculographic signals (EOGs), when monitoring or simulating eye movements. The data are simulated for five different EEG electrode configurations combined with four different EOG electrode configurations. In order to objectively compare correction performance for six algorithms, listed in Table III, we determine the signal to noise ratio of the EEG before and after artifact correction. A score indicating correction performance is derived, and for each EEG configuration the optimal correction algorithm and the optimal number of EOG electrodes are determined. In general, the second-order blind identification correction algorithm in combination with 6 EOG electrodes performs best for all EEG configurations evaluated on the simulated data.
AB - We present a method to quantitatively and objectively compare algorithms for correction of eye movement artifacts in a simulated ongoing electroencephalographic signal (EEG). A
realistic model of the human head is used, together with eye tracker data, to generate a data set in which potentials of ocular and cerebral origin are simulated. This approach bypasses the common problem of brain-potential contaminated electro-oculographic signals (EOGs), when monitoring or simulating eye movements. The data are simulated for five different EEG electrode configurations combined with four different EOG electrode configurations. In order to objectively compare correction performance for six algorithms, listed in Table III, we determine the signal to noise ratio of the EEG before and after artifact correction. A score indicating correction performance is derived, and for each EEG configuration the optimal correction algorithm and the optimal number of EOG electrodes are determined. In general, the second-order blind identification correction algorithm in combination with 6 EOG electrodes performs best for all EEG configurations evaluated on the simulated data.
U2 - 10.1109/TBME.2005.862533
DO - 10.1109/TBME.2005.862533
M3 - Article
C2 - 16485753
SN - 0018-9294
VL - 53
SP - 246
EP - 253
JO - IEEE Transactions on Biomedical Engineering
JF - IEEE Transactions on Biomedical Engineering
IS - 2
ER -